Chapter 13

Rigid Body Motion and Rotational

Dynamics
13.1 Rigid Bodies
A rigid body consists of a group of particles whose separations are all xed in magnitude. Six
independent coordinates are required to completely specify the position and orientation of a
rigid body. For example, the location of the rst particle is specied by three coordinates. A
second particle requires only two coordinates since the distance to the rst is xed. Finally,
a third particle requires only one coordinate, since its distance to the rst two particles
is xed (think about the intersection of two spheres). The positions of all the remaining
particles are then determined by their distances from the rst three. Usually, one takes
these six coordinates to be the center-of-mass position R = (X, Y, Z) and three angles
specifying the orientation of the body (e.g. the Euler angles).
As derived previously, the equations of motion are
P =

i
m
i
r
i
,

P = F
(ext)
(13.1)
L =

i
m
i
r
i
r
i
,

L = N
(ext)
. (13.2)
These equations determine the motion of a rigid body.
13.1.1 Examples of rigid bodies
Our rst example of a rigid body is of a wheel rolling with constant angular velocity

= ,
and without slipping, This is shown in Fig. 13.1. The no-slip condition is dx = Rd, so
x = V
CM
= R. The velocity of a point within the wheel is
v = V
CM
+ r , (13.3)
1
2 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.1: A wheel rolling to the right without slipping.
where r is measured from the center of the disk. The velocity of a point on the surface is
then given by v = R
_
x + r).
As a second example, consider a bicycle wheel of mass M and radius R axed to a
light, rm rod of length d, as shown in Fig. 13.2. Assuming L lies in the (x, y) plane,
one computes the gravitational torque N = r (Mg) = Mgd

. The angular momentum
vector then rotates with angular frequency

. Thus,
d =
dL
L
=

=
Mgd
L
. (13.4)
But L = MR
2
, so the precession frequency is

p
=

=
gd
R
2
. (13.5)
For R = d = 30 cm and /2 = 200 rpm, nd
p
/2 15 rpm. Note that we have here
ignored the contribution to L from the precession itself, which lies along z, resulting in the
nutation of the wheel. This is justied if L
p
/L = (d
2
/R
2
) (
p
/) 1.
13.2 The Inertia Tensor
Suppose rst that a point within the body itself is xed. This eliminates the translational
degrees of freedom from consideration. We now have
_
dr
dt
_
inertial
= r , (13.6)
13.2. THE INERTIA TENSOR 3
Figure 13.2: Precession of a spinning bicycle wheel.
since r
body
= 0. The kinetic energy is then
T =
1
2

i
m
i
_
dr
i
dt
_
2
inertial
=
1
2

i
m
i
( r
i
) ( r
i
)
=
1
2

i
m
i
_

2
r
2
i
( r
i
)
2
_

1
2
I

, (13.7)
where

is the component of along the body-xed axis e

. The quantity I

is the
inertia tensor,
I

i
m
i
_
r
2
i

r
i,
r
i,
_
(13.8)
=
_
d
d
r (r)
_
r
2

_
(continuous media) . (13.9)
The angular momentum is
L =

i
m
i
r
i

_
dr
i
dt
_
inertial
=

i
m
i
r
i
( r
i
) = I

. (13.10)
The diagonal elements of I

are called the moments of inertia, while the o-diagonal

elements are called the products of inertia.
4 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
13.2.1 Coordinate transformations
Consider the basis transformation
e

. (13.11)
We demand e

, which means O(d) is an orthogonal matrix, i.e.

t
=
1
.
Thus the inverse transformation is e

=
t

. Consider next a general vector A = A

.
Expressed in terms of the new basis e

, we have
A = A

..

=
A

..

(13.12)
Thus, the components of A transform as A

. This is true for any vector.

Under a rotation, the density (r) must satisfy

(r

) = (r). This is the transformation

rule for scalars. The inertia tensor therefore obeys
I

=
_
d
3
r

(r

)
_
r

_
=
_
d
3
r (r)
_
r
2

__

_
_
=

. (13.13)
I.e. I

= I
t
, the transformation rule for tensors. The angular frequency is a vector, so

. The angular momentum L also transforms as a vector. The kinetic energy

is T =
1
2

t
I , which transforms as a scalar.
13.2.2 The case of no xed point
If there is no xed point, we can let r

denote the distance from the center-of-mass (CM),

which will serve as the instantaneous origin in the body-xed frame. We then adopt the
notation where R is the CM position of the rotating body, as observed in an inertial frame,
and is computed from the expression
R =
1
M

i
m
i

i
=
1
M
_
d
3
r (r) , (13.14)
where the total mass is of course
M =

i
m
i
=
_
d
3
r (r) . (13.15)
The kinetic energy and angular momentum are then
T =
1
2
M

R
2
+
1
2
I

(13.16)
L

MR

+I

, (13.17)
where I

is given in eqs. 13.8 and 13.9, where the origin is the CM.
13.3. PARALLEL AXIS THEOREM 5
Figure 13.3: Application of the parallel axis theorem to a cylindrically symmetric mass
distribution.
13.3 Parallel Axis Theorem
Suppose I

is given in a body-xed frame. If we displace the origin in the body-xed frame

by d, then let I

(d) be the inertial tensor with respect to the new origin. If, relative to
the origin at 0 a mass element lies at position r, then relative to an origin at d it will lie at
r d. We then have
I

(d) =

i
m
i
_
(r
2
i
2d r
i
+d
2
)

(r
i,
d

)(r
i,
d

)
_
. (13.18)
If r
i
is measured with respect to the CM, then

i
m
i
r
i
= 0 (13.19)
and
I

(d) = I

(0) +M
_
d
2

_
, (13.20)
a result known as the parallel axis theorem.
As an example of the theorem, consider the situation depicted in Fig. 13.3, where a
cylindrically symmetric mass distribution is rotated about is symmetry axis, and about an
axis tangent to its side. The component I
zz
of the inertia tensor is easily computed when
the origin lies along the symmetry axis:
I
zz
=
_
d
3
r (r) (r
2
z
2
) = L 2
a
_
0
dr

r
3

=

2
La
4
=
1
2
Ma
2
, (13.21)
6 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
where M = a
2
L is the total mass. If we compute I
zz
about a vertical axis which is
tangent to the cylinder, the parallel axis theorem tells us that
I

zz
= I
zz
+Ma
2
=
3
2
Ma
2
. (13.22)
Doing this calculation by explicit integration of
_
dmr
2

would be tedious!
13.3.1 Example
Problem: Compute the CM and the inertia tensor for the planar right triangle of Fig.
13.4, assuming it to be of uniform two-dimensional mass density .
Solution: The total mass is M =
1
2
ab. The x-coordinate of the CM is then
X =
1
M
a
_
0
dx
b(1
x
a
)
_
0
dy x =

M
a
_
0
dxb
_
1
x
a
_
x
=
a
2
b
M
1
_
0
duu(1 u) =
a
2
b
6M
=
1
3
a . (13.23)
Clearly we must then have Y =
1
3
b, which may be veried by explicit integration.
We now compute the inertia tensor, with the origin at (0, 0, 0). Since the gure is planar,
z = 0 everywhere, hence I
xz
= I
zx
= 0, I
yz
= I
zy
= 0, and also I
zz
= I
xx
+ I
yy
. We now
compute the remaining independent elements:
I
xx
=
a
_
0
dx
b(1
x
a
)
_
0
dy y
2
=
a
_
0
dx
1
3
b
3
_
1
x
a
_
3
=
1
3
ab
3
1
_
0
du(1 u)
3
=
1
12
ab
3
=
1
6
Mb
2
(13.24)
and
I
xy
=
a
_
0
dx
b(1
x
a
)
_
0
dy xy =
1
2
b
2
a
_
0
dxx
_
1
x
a
_
2
=
1
2
a
2
b
2
1
_
0
duu(1 u)
2
=
1
24
a
2
b
2
=
1
12
Mab . (13.25)
Thus,
I =
M
6
_
_
b
2

1
2
ab 0

1
2
ab a
2
0
0 0 a
2
+b
2
_
_
. (13.26)
13.3. PARALLEL AXIS THEOREM 7
Figure 13.4: A planar mass distribution in the shape of a triangle.
Suppose we wanted the inertia tensor relative in a coordinate system where the CM lies
at the origin. What we computed in eqn. 13.26 is I(d), with d =
a
3
x
b
3
y. Thus,
d
2

=
1
9
_
_
b
2
ab 0
ab a
2
0
0 0 a
2
+b
2
_
_
. (13.27)
Since
I(d) = I
CM
+M
_
d
2

_
, (13.28)
we have that
I
CM
= I(d) M
_
d
2

_
(13.29)
=
M
18
_
_
b
2 1
2
ab 0
1
2
ab a
2
0
0 0 a
2
+b
2
_
_
. (13.30)
13.3.2 General planar mass distribution
For a general planar mass distribution,
(x, y, z) = (x, y) (z) , (13.31)
which is conned to the plane z = 0, we have
I
xx
=
_
dx
_
dy (x, y) y
2
(13.32)
I
yy
=
_
dx
_
dy (x, y) x
2
(13.33)
I
xy
=
_
dx
_
dy (x, y) xy (13.34)
and I
zz
= I
xx
+I
yy
, regardless of the two-dimensional mass distribution (x, y).
8 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
13.4 Principal Axes of Inertia
We found that an orthogonal transformation to a new set of axes e

entails
I

= I
t
for the inertia tensor. Since I = I
t
is manifestly a symmetric matrix, it can
be brought to diagonal form by such an orthogonal transformation. To nd , follow this
recipe:
1. Find the diagonal elements of I

by setting P() = 0, where

P() = det
_
1 I
_
, (13.35)
is the characteristic polynomial for I, and 1 is the unit matrix.
2. For each eigenvalue
a
, solve the d equations

=
a

. (13.36)
Here,
a

is the
th
component of the a
th
eigenvector. Since ( 1 I) is degenerate,
these equations are linearly dependent, which means that the rst d 1 components
may be determined in terms of the d
th
component.
3. Because I = I
t
, eigenvectors corresponding to dierent eigenvalues are orthogonal.
In cases of degeneracy, the eigenvectors may be chosen to be orthogonal, e.g. via the
Gram-Schmidt procedure.
4. Due to the underdetermined aspect to step 2, we may choose an arbitrary normaliza-
tion for each eigenvector. It is conventional to choose the eigenvectors to be orthonor-
mal:

=
ab
.
5. The matrix is explicitly given by
a
=
a

, the matrix whose row vectors are the

eigenvectors
a
. Of course
t
is then the corresponding matrix of column vectors.
6. The eigenvectors form a complete basis. The resolution of unity may be expressed as

. (13.37)
As an example, consider the inertia tensor for a general planar mass distribution, which
is of the form
I =
_
_
I
xx
I
xy
0
I
yx
I
yy
0
0 0 I
zz
_
_
, (13.38)
where I
yx
= I
xy
and I
zz
= I
xx
+I
yy
. Dene
A =
1
2
_
I
xx
+I
yy
_
(13.39)
B =
_
1
4
_
I
xx
I
yy
_
2
+I
2
xy
(13.40)
= tan
1
_
2I
xy
I
xx
I
yy
_
, (13.41)
13.5. EULERS EQUATIONS 9
so that
I =
_
_
A+Bcos Bsin 0
Bsin ABcos 0
0 0 2A
_
_
, (13.42)
The characteristic polynomial is found to be
P() = ( 2A)
_
( A)
2
B
2
_
, (13.43)
which gives
1
= A,
2,3
= AB. The corresponding normalized eigenvectors are

1
=
_
_
0
0
1
_
_
,
2
=
_
_
cos
1
2

sin
1
2

0
_
_
,
3
=
_
_
sin
1
2

cos
1
2

0
_
_
(13.44)
and therefore
=
_
_
0 0 1
cos
1
2
sin
1
2
0
sin
1
2
cos
1
2
0
_
_
. (13.45)
13.5 Eulers Equations
Let us now choose our coordinate axes to be the principal axes of inertia, with the CM at
the origin. We may then write
=
_
_

3
_
_
, I =
_
_
I
1
0 0
0 I
2
0
0 0 I
3
_
_
= L =
_
_
I
1

1
I
2

2
I
3

3
_
_
. (13.46)
The equations of motion are
N
ext
=
_
dL
dt
_
inertial
=
_
dL
dt
_
body
+ L
= I + (I ) .
Thus, we arrive at Eulers equations:
I
1

1
= (I
2
I
3
)
2

3
+N
ext
1
(13.47)
I
2

2
= (I
3
I
1
)
3

1
+N
ext
2
(13.48)
I
3

3
= (I
1
I
2
)
1

2
+N
ext
3
. (13.49)
These are coupled and nonlinear. Also note the fact that the external torque must be
evaluated along body-xed principal axes. We can however make progress in the case
10 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.5: Wobbling of a torque-free symmetric top.
where N
ext
= 0, i.e. when there are no external torques. This is true for a body in free
space, or in a uniform gravitational eld. In the latter case,
N
ext
=

i
r
i
(m
i
g) =
_

i
m
i
r
i
_
g , (13.50)
where g is the uniform gravitational acceleration. In a body-xed frame whose origin is the
CM, we have

i
m
i
r
i
= 0, and the external torque vanishes!
Precession of torque-free symmetric tops: Consider a body which has a symme-
try axis e
3
. This guarantees I
1
= I
2
, but in general we still have I
1
,= I
3
. In the absence
of external torques, the last of Eulers equations says
3
= 0, so
3
is a constant. The
remaining two equations are then

1
=
_
I
1
I
3
I
1
_

2
,
2
=
_
I
3
I
1
I
1
_

1
. (13.51)
I.e.
1
=
2
and
2
= +
1
, with
=
_
I
3
I
1
I
1
_

3
, (13.52)
which are the equations of a harmonic oscillator. The solution is easily obtained:

1
(t) =

cos
_
t +
_
,
2
(t) =

sin
_
t +
_
,
3
(t) =
3
, (13.53)
where

and are constants of integration, and where [[ = (

2

+
2
3
)
1/2
. This motion is
sketched in Fig. 13.5. Note that the perpendicular components of oscillate harmonically,
and that the angle makes with respect to e
3
is = tan
1
(

/
3
).
For the earth, (I
3
I
1
)/I
1

1
305
, so
3
, and /305, yielding a precession period
of 305 days, or roughly 10 months. Astronomical observations reveal such a precession,
13.5. EULERS EQUATIONS 11
known as the Chandler wobble. For the earth, the precession angle is
Chandler
6 10
7
rad, which means that the North Pole moves by about 4 meters during the wobble. The
Chandler wobble has a period of about 14 months, so the nave prediction of 305 days is
o by a substantial amount. This discrepancy is attributed to the mechanical properties of
the earth: elasticity and uidity. The earth is not solid!
1
Asymmetric tops: Next, consider the torque-free motion of an asymmetric top, where
I
1
,= I
2
,= I
3
,= I
1
. Unlike the symmetric case, there is no conserved component of . True,
we can invoke conservation of energy and angular momentum,
E =
1
2
I
1

2
1
+
1
2
I
2

2
2
+
1
2
I
3

2
3
(13.54)
L
2
= I
2
1

2
1
+I
2
2

2
2
+I
2
3

2
3
, (13.55)
and, in principle, solve for
1
and
2
in terms of
3
, and then invoke Eulers equations
(which must honor these conservation laws). However, the nonlinearity greatly complicates
matters and in general this approach is a dead end.
We can, however, nd a particular solution quite easily one in which the rotation is
about a single axis. Thus,
1
=
2
= 0 and
3
=
0
is indeed a solution for all time,
according to Eulers equations. Let us now perturb about this solution, to explore its
stability. We write
=
0
e
3
+ , (13.56)
and we invoke Eulers equations, linearizing by dropping terms quadratic in . This yield
I
1

1
= (I
2
I
3
)
0

2
+O(
2

3
) (13.57)
I
2

2
= (I
3
I
1
)
0

1
+O(
3

1
) (13.58)
I
3

3
= 0 +O(
1

2
) . (13.59)
Taking the time derivative of the rst equation and invoking the second, and vice versa,
yields

1
=
2

1
,
2
=
2

2
, (13.60)
with

2
=
(I
3
I
2
)(I
3
I
1
)
I
1
I
2

2
0
. (13.61)
The solution is then
1
(t) = C cos(t +).
If
2
> 0, then is real, and the deviation results in a harmonic precession. This occurs
if I
3
is either the largest or the smallest of the moments of inertia. If, however, I
3
is the
middle moment, then
2
< 0, and is purely imaginary. The perturbation will in general
increase exponentially with time, which means that the initial solution to Eulers equations
is unstable with respect to small perturbations. This result can be vividly realized using a
tennis racket, and sometimes goes by the name of the tennis racket theorem.
1
The earth is a layered like a Mozartkugel, with a solid outer shell, an inner uid shell, and a solid (iron)
core.
12 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
13.5.1 Example
PROBLEM: A unsuspecting solid spherical planet of mass M
0
rotates with angular velocity

0
. Suddenly, a giant asteroid of mass M
0
smashes into and sticks to the planet at a
location which is at polar angle relative to the initial rotational axis. The new mass
distribution is no longer spherically symmetric, and the rotational axis will precess. Recall
Eulers equation
dL
dt
+ L = N
ext
(13.62)
for rotations in a body-xed frame.
(a) What is the new inertia tensor I

along principle center-of-mass frame axes? Dont

forget that the CM is no longer at the center of the sphere! Recall I =
2
5
MR
2
for a solid
sphere.
(b) What is the period of precession of the rotational axis in terms of the original length
of the day 2/
0
?
SOLUTION: Lets choose body-xed axes with z pointing from the center of the planet to
the smoldering asteroid. The CM lies a distance
d =
M
0
R +M
0
0
(1 +)M
0
=

1 +
R (13.63)
from the center of the sphere. Thus, relative to the center of the sphere, we have
I =
2
5
M
0
R
2
_
_
1 0 0
0 1 0
0 0 1
_
_
+M
0
R
2
_
_
1 0 0
0 1 0
0 0 0
_
_
. (13.64)
Now we shift to a frame with the CM at the origin, using the parallel axis theorem,
I

(d) = I
CM

+M
_
d
2

_
. (13.65)
Thus, with d = d z,
I
CM

=
2
5
M
0
R
2
_
_
1 0 0
0 1 0
0 0 1
_
_
+M
0
R
2
_
_
1 0 0
0 1 0
0 0 0
_
_
(1 +)M
0
d
2
_
_
1 0 0
0 1 0
0 0 0
_
_
(13.66)
= M
0
R
2
_
_
2
5
+

1+
0 0
0
2
5
+

1+
0
0 0
2
5
_
_
. (13.67)
13.6. EULERS ANGLES 13
In the absence of external torques, Eulers equations along principal axes read
I
1
d
1
dt
= (I
2
I
3
)
2

3
I
2
d
2
dt
= (I
3
I
1
)
3

1
I
3
d
3
dt
= (I
1
I
2
)
1

2
(13.68)
Since I
1
= I
2
,
3
(t) =
3
(0) =
0
cos is a constant. We then obtain
1
=
2
, and

2
=
1
, with
=
I
2
I
3
I
1

3
=
5
7 + 2

3
. (13.69)
The period of precession in units of the pre-cataclysmic day is

T
=

=
7 + 2
5cos
. (13.70)
13.6 Eulers Angles
In d dimensions, an orthogonal matrix O(d) has
1
2
d(d 1) independent parameters.
To see this, consider the constraint
t
= 1. The matrix
t
is manifestly symmetric,
so it has
1
2
d(d + 1) independent entries (e.g. on the diagonal and above the diagonal).
This amounts to
1
2
d(d + 1) constraints on the d
2
components of , resulting in
1
2
d(d 1)
freedoms. Thus, in d = 3 rotations are specied by three parameters. The Euler angles
, , provide one such convenient parameterization.
A general rotation (, , ) is built up in three steps. We start with an orthonormal
triad e
0

of body-xed axes. The rst step is a rotation by an angle about e

0
3
:
e

_
e
0
3
,
_
e
0

,
_
e
0
3
,
_
=
_
_
cos sin 0
sin cos 0
0 0 1
_
_
(13.71)
This step is shown in panel (a) of Fig. 13.6. The second step is a rotation by about the
new axis e

1
:
e

_
e

1
,
_
e

,
_
e

1
,
_
=
_
_
1 0 0
0 cos sin
0 sin cos
_
_
(13.72)
This step is shown in panel (b) of Fig. 13.6. The third and nal step is a rotation by
about the new axis e

3
:
e

_
e

3
,
_
e

,
_
e

3
,
_
=
_
_
cos sin 0
sin cos 0
0 0 1
_
_
(13.73)
14 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.6: A general rotation, dened in terms of the Euler angles , , . Three
successive steps of the transformation are shown.
This step is shown in panel (c) of Fig. 13.6. Putting this all together,
(, , ) =
_
e

3
,
_

_
e

1
,
_

_
e
0
3
,
_
(13.74)
=
_
_
cos sin 0
sin cos 0
0 0 1
_
_
_
_
1 0 0
0 cos sin
0 sin cos
_
_
_
_
cos sin 0
sin cos 0
0 0 1
_
_
=
_
_
cos cos sin cos sin cos sin + sin cos cos sin sin
sin cos cos cos sin sin sin + cos cos cos cos sin
sin sin sin cos cos
_
_
.
Next, wed like to relate the components

= e

(with e

) of the rotation in
the body-xed frame to the derivatives

,

, and

. To do this, we write
=

e

+

e

+

e

, (13.75)
where
e
0
3
= e

= sin sin e
1
+ sin cos e
2
+ cos e
3
(13.76)
e

= cos e
1
sin e
2
(line of nodes) (13.77)
e

= e
3
. (13.78)
13.6. EULERS ANGLES 15
This gives

1
= e
1
=

sin sin +

cos (13.79)

2
= e
2
=

sin cos

sin (13.80)

3
= e
3
=

cos +

. (13.81)
Note that

precession ,

nutation ,

axial rotation . (13.82)
The general form of the kinetic energy is then
T =
1
2
I
1
_

sin sin +

cos
_
2
+
1
2
I
2
_

sin cos

sin
_
2
+
1
2
I
3
_

cos +

_
2
. (13.83)
Note that
L = p

+p

+p

, (13.84)
which may be veried by explicit computation.
13.6.1 Torque-free symmetric top
A body falling in a gravitational eld experiences no net torque about its CM:
N
ext
=

i
r
i
(m
i
g) = g

i
m
i
r
i
= 0 . (13.85)
For a symmetric top with I
1
= I
2
, we have
T =
1
2
I
1
_

2
+

2
sin
2

_
+
1
2
I
3
_

cos +

_
2
. (13.86)
The potential is cyclic in the Euler angles, hence the equations of motion are
d
dt
T
(

,

,

)
=
T
(, , )
. (13.87)
Since and are cyclic in T, their conjugate momenta are conserved:
p

=
L

= I
1

sin
2
+I
3
(

cos +

) cos (13.88)
p

=
L

= I
3
(

cos +

) . (13.89)
Note that p

= I
3

3
, hence
3
is constant, as we have already seen.
To solve for the motion, we rst note that L is conserved in the inertial frame. We are
therefore permitted to dene

L = e
0
3
= e

. Thus, p

= L. Since e

= cos , we have
16 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.7: A dreidl is a symmetric top. The four-fold symmetry axis guarantees I
1
= I
2
.
The blue diamond represents the center-of-mass.
that p

= L e

= Lcos . Finally, e

= 0, which means p

= L e

= 0. From the
equations of motion,
p

= I
1

=
_
I
1

cos p

_

sin , (13.90)
hence we must have

= 0 ,

=
p

I
1
cos
. (13.91)
Note that

= 0 follows from conservation of p

= Lcos . From the equation for p

, we
may now conclude

=
p

I
3

I
1
=
_
I
3
I
1
I
3
_

3
, (13.92)
which recapitulates (13.52), with

= .
13.6.2 Symmetric top with one point xed
Consider the case of a symmetric top with one point xed, as depicted in Fig. 13.7. The
Lagrangian is
L =
1
2
I
1
_

2
+

2
sin
2

_
+
1
2
I
3
_

cos +

_
2
Mg cos . (13.93)
Here, is the distance from the xed point to the CM, and the inertia tensor is dened along
principal axes whose origin lies at the xed point (not the CM!). Gravity now supplies a
torque, but as in the torque-free case, the Lagrangian is still cyclic in and , so
p

= (I
1
sin
2
+I
3
cos
2
)

+I
3
cos

(13.94)
p

= I
3
cos

+I
3

(13.95)
13.6. EULERS ANGLES 17
Figure 13.8: The eective potential of eq. 13.102.
are each conserved. We can invert these relations to obtain

and

in terms of p

, p

, :

=
p

cos
I
1
sin
2

,

=
p

I
3

(p

cos ) cos
I
1
sin
2

. (13.96)
In addition, since L/t = 0, the total energy is conserved:
E = T +U =
1
2
I
1

2
+
U
e
()
..
(p

cos )
2
2I
1
sin
2

+
p
2

2I
3
+Mg cos , (13.97)
where the term under the brace is the eective potential U
e
().
The problem thus reduces to the one-dimensional dynamics of (t), i.e.
I
1

=
U
e

, (13.98)
with
U
e
() =
(p

cos )
2
2I
1
sin
2

+
p
2

2I
3
+Mg cos . (13.99)
Using energy conservation, we may write
dt =
_
I
1
2
d
_
E U
e
()
. (13.100)
and thus the problem is reduced to quadratures:
t() = t(
0
)
_
I
1
2

0
d
1
_
E U
e
()
. (13.101)
18 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.9: Precession and nutation of the symmetry axis of a symmetric top.
We can gain physical insight into the motion by examining the shape of the eective
potential,
U
e
() =
(p

cos )
2
2I
1
sin
2

+Mg cos +
p
2

2I
3
, (13.102)
over the interval [0, ]. Clearly U
e
(0) = U
e
() = , so the motion must be bounded.
What is not yet clear, but what is nonetheless revealed by some additional analysis, is that
U
e
() has a single minimum on this interval, at =
0
. The turning points for the motion
are at =
a
and =
b
, where U
e
(
a
) = U
e
(
b
) = E. Clearly if we expand about
0
and
write =
0
+, the motion will be harmonic, with
(t) =
0
cos(t +) , =

e
(
0
)
I
1
. (13.103)
To prove that U
e
() has these features, let us dene u cos . Then u =

sin, and
from E =
1
2
I
1

2
+U
e
() we derive
u
2
=
_
2E
I
1

p
2

I
1
I
3
_
(1 u
2
)
2Mg
I
1
(1 u
2
) u
_
p

u
I
1
_
2
f(u) . (13.104)
The turning points occur at f(u) = 0. The function f(u) is cubic, and the coecient of the
cubic term is 2Mg/I
1
, which is positive. Clearly f(u = 1) = (p

)
2
/I
2
1
is negative,
so there must be at least one solution to f(u) = 0 on the interval u (1, ). Clearly there
can be at most three real roots for f(u), since the function is cubic in u, hence there are at
most two turning points on the interval u [1, 1]. Thus, U
e
() has the form depicted in
g. 13.8.
To apprehend the full motion of the top in an inertial frame, let us follow the symmetry
axis e
3
:
e
3
= sin sin e
0
1
sin cos e
0
2
+ cos e
0
3
. (13.105)
Once we know (t) and (t) were done. The motion (t) is described above: oscillates
between turning points at
a
and
b
. As for (t), we have already derived the result

=
p

cos
I
1
sin
2

. (13.106)
13.7. ROLLING AND SKIDDING MOTION OF REAL TOPS 19
Figure 13.10: A top with a peg end. The frictional forces f and f
skid
are shown. When the
top rolls without skidding, f
skid
= 0.
Thus, if p

> p

cos
a
, then

will remain positive throughout the motion. If, on the other
hand, we have
p

cos
b
< p

< p

cos
a
, (13.107)
then

changes sign at an angle

= cos
1
_
p

/p

_
. The motion is depicted in Fig. 13.9.
An extensive discussion of this problem is given in H. Goldstein, Classical Mechanics.
13.7 Rolling and Skidding Motion of Real Tops
The material in this section is based on the corresponding sections from V. Barger and
M. Olsson, Classical Mechanics: A Modern Perspective. This is an excellent book which
contains many interesting applications and examples.
13.7.1 Rolling tops
In most tops, the point of contact rolls or skids along the surface. Consider the peg end top
of Fig. 13.10, executing a circular rolling motion, as sketched in Fig. 13.11. There are three
components to the force acting on the top: gravity, the normal force from the surface, and
friction. The frictional force is perpendicular to the CM velocity, and results in centripetal
acceleration of the top:
f = M
2
Mg , (13.108)
where is the frequency of the CM motion and is the coecient of friction. If the above
inequality is violated, the top starts to slip.
The frictional and normal forces combine to produce a torque N = Mg sin f cos
20 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.11: Circular rolling motion of the peg top.
about the CM
2
. This torque is tangent to the circular path of the CM, and causes L to
precess. We assume that the top is spinning rapidly, so that L very nearly points along
the symmetry axis of the top itself. (As well see, this is true for slow precession but not
for fast precession, where the precession frequency is proportional to
3
.) The precession is
then governed by the equation
N = Mg sin f cos
=

I
3

3
sin , (13.109)
where e
3
is the instantaneous symmetry axis of the top. Substituting f = M
2
,
Mg
I
3

3
_
1

2

g
ctn
_
= , (13.110)
which is a quadratic equation for . We supplement this with the no slip condition,

3
=
_
+ sin
_
, (13.111)
resulting in two equations for the two unknowns and .
Substituting for () and solving for , we obtain
=
I
3

3
2M
2
cos
_
1 +
Mg
I
3
ctn

_
1 +
Mg
I
3
ctn
_
2

4M
2
I
3

Mg
I
3

2
3
_
. (13.112)
This in order to have a real solution we must have

3

2M
2
sin
I
3
sin +Mg cos
_
g

. (13.113)
2
Gravity of course produces no net torque about the CM.
13.7. ROLLING AND SKIDDING MOTION OF REAL TOPS 21
If the inequality is satised, there are two possible solutions for , corresponding to fast
and slow precession. Usually one observes slow precession. Note that it is possible that
< 0, in which case the CM and the peg end lie on opposite sides of a circle from each
other.
13.7.2 Skidding tops
A skidding top experiences a frictional force which opposes the skidding velocity, until
v
skid
= 0 and a pure rolling motion sets in. This force provides a torque which makes the
top rise:

=
N
skid
L
=
Mg
I
3

3
. (13.114)
Suppose 0, in which case + sin = 0, from eqn. 13.111, and the point of contact
remains xed. Now recall the eective potential for a symmetric top with one point xed:
U
e
() =
(p

cos )
2
2I
1
sin
2

+
p
2

2I
3
+Mg cos . (13.115)
We demand U

e
(
0
) = 0, which yields
cos
0

2
p

sin
2

0
+MgI
1
sin
4

0
= 0 , (13.116)
where
p

cos
0
= I
1
sin
2

. (13.117)
Solving the quadratic equation for , we nd

=
I
3

3
2I
1
cos
0
_
1

1
4MgI
1
cos
0
I
2
3

2
3
_
. (13.118)
This is simply a recapitulation of eqn. 13.112, with = 0 and with M
2
replaced by I
1
.
Note I
1
= M
2
by the parallel axis theorem if I
CM
1
= 0. But to the extent that I
CM
1
,= 0, our
treatment of the peg top was incorrect. It turns out to be OK, however, if the precession is
slow, i.e. if /
3
1.
On a level surface, cos
0
> 0, and therefore we must have

3

2
I
3
_
MgI
1
cos
0
. (13.119)
Thus, if the top spins too slowly, it cannot maintain precession. Eqn. 13.118 says that there
are two possible precession frequencies. When
3
is large, we have

slow
=
Mg
I
3

3
+O(
1
3
) ,

fast
=
I
3

3
I
1
cos
0
+O(
3
3
) . (13.120)
Again, one usually observes slow precession.
22 CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.12: The tippie-top behaves in a counterintuitive way. Once started spinning with
the peg end up, the peg axis rotates downward. Eventually the peg scrapes the surface and
the top rises to the vertical in an inverted orientation.
A top with
3
>
2
I
3

MgI
1
may sleep in the vertical position with
0
= 0. Due to the
constant action of frictional forces,
3
will eventually drop below this value, at which time
the vertical position is no longer stable. The top continues to slow down and eventually
falls.
13.7.3 Tippie-top
A particularly nice example from the Barger and Olsson book is that of the tippie-top, a
truncated sphere with a peg end, sketched in Fig. 13.12 The CM is close to the center
of curvature, which means that there is almost no gravitational torque acting on the top.
The frictional force f opposes slipping, but as the top spins f rotates with it, and hence
the time-averaged frictional force f) 0 has almost no eect on the motion of the CM.
A similar argument shows that the frictional torque, which is nearly horizontal, also time
averages to zero:
_
dL
dt
_
inertial
0 . (13.121)
In the body-xed frame, however, N is roughly constant, with magnitude N MgR,
where R is the radius of curvature and the coecient of sliding friction. Now we invoke
N =
dL
dt

body
+ L . (13.122)
The second term on the RHS is very small, because the tippie-top is almost spherical, hence
13.7. ROLLING AND SKIDDING MOTION OF REAL TOPS 23
inertia tensor is very nearly diagonal, and this means
L I = 0 . (13.123)
Thus,

L
body
N, and taking the dot product of this equation with the unit vector

k, we
obtain
N sin =

k N =
d
dt
_

k L
body
_
= Lsin

. (13.124)
Thus,

=
N
L

MgR
I
. (13.125)
Once the stem scrapes the table, the tippie-top rises to the vertical just like any other rising
top.